Background

"Language exerts hidden power, like a moon on the tides." (Rita Mae Brown, Starting From Scratch, New York: Bantam, 1988)

What we call things matters – reports of medical errors "channel attention, shape interpretations and serve as springboards for action"[1]. Many entities require the reporting of errors, [2–7] and with the passage of the United States Patient Safety and Quality Improvement Act of 2005, it is likely that even more physicians will be asked to identify and report errors[8]. Once a domain primarily of hospitals, the importance of medical errors occurring in the outpatient, primary care setting has become more apparent, [9–15] and physicians are being asked to report errors from this venue as well.

Yet, while physicians are being asked to find, report and reduce medical errors in their practices, we lack a universally understood definition of exactly what is meant by "medical error"[1, 16–19]. Previous primary care studies have demonstrated large differences in the number of errors reported by individual physicians, even within the same practices[12, 20] and research in hospitals has found differences in how health care workers interpret terms like error, incident and event[21, 22]. Individual interpretation of what is an error plays a role in identifying errors and making reports[23, 24]. While collations of some patient safety terms from the literature have been done, [25, 26] how these definitions affect physicians' use of terms like "error" is not clear.

In order to better understand what may affect a physician's understanding of "medical error," we performed a three step process: 1) We systematically collected definitions for medical error found in the medical literature; 2) We surveyed family physicians and family physician medical error "experts" about whether they felt a medical error occurred in a series of common clinical scenarios; and 3) We qualitatively explored both the definitions and the survey findings to see if a model of factors might help explain how physicians make decisions about whether to call something an error.

Methods

Literature search

Using the MESH term Medical Error/classification, we retrieved 216 English language articles from 1985 – October, 2005. All abstracts were reviewed, and 68 articles dealing with "medical error" or "error" were reviewed for definitions. Further articles containing definitions were gleaned from the medical errors literature searching with the MESH term Medical Error and the secondary text phrases "primary care," "family medicine" and "error reporting" (93, 37 and 78 abstracts reviewed, and a total of 16 additional articles reviewed). Definitions that focused exclusively on a subset of error, such as medication or diagnosis, were excluded. In addition, the report of the World Health Organization on a draft comparative glossary of patient safety terms, prepared by the Joint Commission for the Accreditation of Healthcare Organizations was also reviewed[25, 27]. In addition, definitions used in national surveys (United States) and polls were collated and added to the definition list.

Survey development

From the literature[2, 10, 12] and our experience, we devised 5 common scenarios that were representative of preventable problems experienced by family physicians. Due to the frequency and importance of the testing process in primary care, [10, 12, 28, 29] we chose this area in which to concentrate our scenarios, as they were likely to be familiar and real to practicing physicians. These scenarios were then sent to a group of primary care medical error researchers for review and a draft of the survey was pilot tested on 10 local family physicians for refinement prior to the final survey. The final five scenarios are found in table 1. We purposely chose to include scenarios that included variability around a number of factors, including medical conditions, systems and individuals in a practice and knowledge and severity of clinical outcomes. In asking our participants their opinions on the scenarios, we asked whether "an error or mistake had occurred." We used both words because some definitions state that an error is a plan not achieving its goal while a mistake is an incorrect plan,[30] and we wanted to be inclusive of both issues. We did not, however, define either term on the survey, as our intention was to assess agreement about their use as free standing words.

Table 1 Clinical scenarios used in the survey
Full size table

Data collection

We elected to survey both a random sample of active family physician members of the AAFP as well as a selected sample of family physician "experts" that have presented, published or advocated nationally about medical error. We did this in order to look for a possible disconnect between the application of error by these two groups. After approval by our institutional review board, we mailed our survey and a cover letter to a random sample of 1000 active family physician members of the AAFP and by e-mail to 25 family physician experts. All non-responders were sent a second survey 3 weeks later. Demographic and practice data were also collected about each AAFP respondent. Data were entered into an excel database.

Data analysis

Data were reviewed for accuracy, and descriptive statistics were performed using SPSS. The survey was performed for descriptive purposes only, and was not powered for statistical analyses. While the survey did not ask for comments, several participants added hand written comments and these were separated and reviewed. Data sources for the qualitative analysis were collated and included the survey scenarios, the survey results, hand written comments, and the definitions from the literature review. These were analyzed together using the qualitative technique of immersion and crystallization[31, 32]. With this technique, we immersed ourselves in these data, gaining emergent insights. Immersion included reading, re-reading, organizing phrases and segments of written text, model building and discussion. We then participated in a series of discussions in order to explicate theories and synthesize ideas. We then returned to the medical literature and the data sources looking for both corroborating findings and alternative interpretations[31].

Results

Literature review

The most common definition found in the medical literature is that attributed to James Reason and published by the Institute of Medicine (IOM) in To Err is Human, "Failure of a planned action to be completed as intended or use of a wrong plan to achieve an aim"[33]. (Table 2) Because it is so commonly used, we included several variations on it. Interestingly, James Reason originally defined "error" only as the first part – failure of a planned action to be completed as intended. He used the term "mistake" for use of a wrong plan to achieve an aim[30]. However, the IOM report used "error" for both components, defining one as an error of execution and the other as an error of planning[33]. Reason also defined slips as relating to observable actions, associated with attentional failures and lapses as more internal and related to failures of memory[30].

Table 2 Medical error definitions from the medical literature
Full size table

Of note is that several definitions include comments about actual or potential outcomes, for example, "a threat to patient safety,"[34] "potentially negative consequences,"[35] "result in a perceived harm,"[14] or "result in serious harm"[36]. Other definitions, however, focus only on the processes. In addition, some definitions focus on the human components,[37] while others more on systems[38]. Surveys and polls to both the public and physicians tend to define by example and include harmful outcomes in the definition[36, 39].

Survey

Participants

From the AAFP sample of active members, we received 284 usable surveys from 997 valid addresses for a 28.5% response rate. Demographic details of the respondents are found in Table 3. Our sample compares favorably with the total active membership of the AAFP. The AAFP is 32.3% women, and our sample was 35.8%. The age breakdown of the AAFP is almost identical to our age breakdown (30 – 44: AAFP 45.8%, our study 46%; 45–54: AAFP 45%, our study 48%; 60 and over: AAFP 6.8%, our study 6.0%). Our sample had 15% solo practitioners (18% in AAFP) and 25.2% in a multi-specialty practice (21.7% in AAFP)[40]. Twenty three of 25 "experts" responded to the survey (92%), and two-thirds of them were male with an average age of 47 (range 34 – 58).

Table 3 Demographics of participants
Full size table

Error occurrence

One hundred percent of the AAFP respondents believed an "error or mistake" occurred when a physician wrote "normal, file" on an abnormal result (Ms Brown/TSH), while only 47% felt that a dropped blood tube was an error (Mr. White/broken tube). (Table 4) Several physicians believed the broken blood tube was "an accident." Almost 14% of physicians felt they couldn't make a decision about whether an error occurred when confronted with the scenario about missing results (Ms Green/CT results), but 62% did believe there was an error. The experts' opinions mirrored those of the AAFP members. The experts committed to a decision in all the scenarios except the Ms Green/CT results scenario, with 13% being unable to decide if an error occurred.

Table 4 Percent of respondents who believed an error or mistake occurred in the described scenario.
Full size table

Qualitative categorization of "error" decision making

Since all the scenarios described a situation where a "failure of a planned action to be completed as intended or use of a wrong plan to achieve an aim" occurred, we looked for additional explanatory processes to explain the variation in the survey results. We found three areas that we felt could explain physician decision making around "is this event is an error?": Do I know the outcome and is there harm from this event, is this event a common or a rare occurrence and does responsibility for this event lie predominantly with an individual or with the system? (Table 5) We found error definitions that imposed criteria for calling an event an error for all of these explanatory areas, and the findings of the survey supported their use in error decision making. Another area of potential conflict for physicians is that of "reasonable expectations"[10, 41–43]. However, there were no survey scenarios, findings or comments that seemed to address this, and we chose not to include it in our model, but it does warrant further research.

Table 5 Factors associated with assigning error to a scenario as determined by qualitative analysis.
Full size table

A model of the three questions used by physicians in the decision making process is found in figure 1. As physicians ponder calling an event an error, they first make the decision that something went wrong, and then look to these three areas to tip the balance towards or away from calling the event an error. Using a survey scenario, for example, Ms Brown/TSH had harm, is likely a rare occurrence and has clear individual responsibility, and 100% of participants felt it was an error.

Figure 1
figure 1

A model of physician decision making when assessing whether an event should be classified as an error.

Full size image

Discussion

Error and its many synonyms not only appear in medical journals, they are words used frequently in daily conversation. James Reason's definition is widely accepted, frequently cited in the medical literature and encompasses human and system processes[33, 44]. Yet, in attempts to clarify, expand and modify, many others have also defined medical error. (Table 2) After deciding something is "not right," additional questions, even if not explicitly asked, likely figure into the decision making process: Do I know the outcome and is there harm from this event, is this event a common or a rare occurrence and does responsibility for this event lie predominantly with an individual or with the system? (Figure 1)

In both the medical literature and for our participants, outcome seems to influence determinations about error[45, 46]. Considering the outcome as well as the process in making decisions about error is not unusual[11, 17]. As noted, several of the error definitions found in the literature include harm or potential harm in the definition[14, 35, 36, 39, 47–49]. Woods and Cook remark on this confusion by describing three ways in which "error" is used: error as the cause of failure (or poor outcome), as the failure itself or as departure from a standard process[46]. Hindsight and outcome bias describe how knowledge of the outcome affects the decision making about the quality of processes[50, 51]. For example, if we can't find a test result when wanted, but it is normal and doesn't change our management, we define the lost result differently than when a result is abnormal, and now treatment has been significantly delayed[21]. The process of losing the result may be the same, but our "hindsight" of the "outcome" affects how we interpret that process. Our physician respondents may be so used to using the hindsight of outcome to assess a process, that without that information (especially with the missing CT result) it is difficult for them make a decision about just the process. Similar findings in other studies note that uncertain outcomes lead to larger proportions of respondents refusing to make a decision about error[21]. Tamuz and colleagues found that errors detected and corrected by staff in a hospital were felt to be non-events that occurred as a natural part of the work flow, and not error[1, 22]. In the model in figure 1, knowledge of harm tips the balance towards labeling an event as error, but not knowing the outcome leaves the balance unswayed.

Some problems may occur so commonly in practice today, that it is difficult for physicians to perceive these as "errors." For example, there is probably not a practice in existence that draws blood that has not lost, broken or somehow damaged a blood specimen tube. And missing clinical information has recently been documented to occur in approximately 14% of all office visits[29]. The two scenarios we offered physicians of these commonly occurring events, Mr. White/broken tube and Ms Green/CT results, received the least agreement from physicians that an error occurred. Perhaps the fact that these experiences are weekly, if not daily occurrences for many makes it difficult for physicians to acknowledge these as "errors." While these scenarios certainly fit definitions of medical error, [2, 10, 48], these "academic" definitions may seem disconnected from practicing physicians who have emotional and personal responses to words such as error[22, 42, 52]. Events that are infrequent and unexpected are more likely to tip the balance toward error (figure 1), whereas those that are common and expected (like an occasional broken blood tube or missing test results) tip the balance away from error[22, 24].

There may also be differences in how errors are perceived by physicians whether the problem appears to be in the system or due to an individual's action. While three scenarios had strong agreement that an error occurred (Mr. Black/LFT, Mrs. Rose/glc and Ms Brown/TSH), in only one, where the physician misreads an abnormal TSH is there 100% agreement from all the participants. Traditionally, when errors occurred, the standard response was to "blame and shame"[33, 53]. A responsible person is most clearly identified in the Ms Brown/TSH scenario, and may be part of the reason why all physicians identified an error in this scenario. Physicians may more easily identify errors where the decision and action of an individual are at fault, rather than where the system fell down. A qualitative, hospital based study previously found that complex system errors were more likely to be called "practice variances" or "suboptimal outcomes" rather than error[21]. Reason and others have addressed this complexity of both system problems and human actions by describing "latent" errors (the underlying system) and "active" errors (the human actions) [30, 44, 54, 55]. But the emotional aspects of feeling responsible may tip the balance toward deciding an event is an error more than an understanding of complex systems.

There are several limitations to this study. The literature search was systematic, but not exhaustive. Non English language articles were not reviewed, and the medical errors literature is too large to review the body of all articles to see if a definition of error is proposed. However, our review did reveal a broad spectrum of error definition, not previously collated and published. Our survey response rate of 28.5% from the AAFP solicitation is low, and certainly limits the generalizability of the results. Those who chose to respond may differ in their opinions about the issues under study. However, we used the survey qualitatively to illuminate and illustrate potential deficiencies in error definitions. The scenarios we devised came from our clinical experience and our experience in researching testing process errors in family physician offices[11, 12, 20]. Although the scenarios were reviewed for face validity and pilot tested, there still may have been some unclear sections of the scenarios, leading to responses that we might have misinterpreted. However, this method mimics the practical application of error decision making, and has been used successfully to study patient safety and errors[21, 56, 57]. Use of the surveys did not allow us to discover if factors such as knowledge of harm were being used inappropriately by our participants in making decisions about error, and this is an area for further research.

Conclusion

Physicians are being asked to make reports of errors, and this is likely to increase in the future. Error reports can be essential to determining the focus of patient safety attention and interventions[1]. Further research is needed to better understand how physicians make decisions about calling an event an error. We generated a model from this study that proposes that three additional elements are important in making a decision about whether an event is an error: Do I know the outcome and is there harm from this event, is this event a common or a rare occurrence and does responsibility for this event lie predominantly with an individual or with the system? The relative importance of each of these areas, and their interrelationships need to be confirmed by further research, including both qualitative and quantitative studies. As we better understand how family physicians use the word "error" then the reports they make will be even more useful as springboards for action.